Re: [ga] is ICANN or is ICANN not? - "CS" fate
Roberto Gaetano wrote: Elisabeth Porteneuve wrote: This is one of those situations that Lessig wrote about, how running code coerces choices. Domain names, at the protocol level, do not have dots. And are composed of a sequence of length/value pairs. In the protocol the root comes at the end of the sequence, the TLD would be the penultimate item in the sequence, etc. Thus in www.cavebear.com, the labels would be, in sequence, 3/www, 8/cavebear, 3/com, 0/(root) What we see as "www.cavebear.com" is largely a textual convention for human consumption. It is a convention that, however, is embedded into zone files and URLs/URI's and has thus become effectively immutable. Nearly all software resolves names by calling a function called "gethostbyname". (There are variations, but for our purposes we need not be concerned with those variations.) That code encloses and encapsulates the interpretation of the human-visible sequence of domain name labels. There is no context information, no locale, that is carred along with domain names to indicate whether the textual sequence of labels is the normal ...TLD.root format or the root.TLD... format. There is no way for gethostbyname() to be smart, the best it could do would be a heuristic reordering in case a primary lookup fails. When the UK deviate format was live there was a great deal of confusion both by people and in software. Those were the days before DNS and thus the cure was largely to do name lookups in the local hostname file using the entire textual string, and populating the hostname file with both variants for the UK deviant names or having software try to reverse the name if the normal scan failed. It was a downright ugly situation and to some extent did cause an impedance mismatch when sending email to/from the UK. --karl--
|